System of passenger tracking in the terminal IS-PAC

Control access of one passenger is a simple task but handling passenger traffic is quite a different matter.

IS-PAC (Passenger Access Control) is designed for passengers’ access control and movement tracking in the terminal, profiling and volume assessment of passenger traffic.

IS-PAC system is complemented with HD cameras what allows the system to solve the following main tasks with high precision:

  • Passengers’ access control to areas with restricted access;
  • Passenger traffic control via points of pre-flight inspection to sterile areas only;
  • Automatic profiling of passengers’ behaviour;
  • Tracking of forgotten baggage in terminals and airport surrounding area;
  • Detection of conflicts and fights in the terminal.

Non-standard behaviour profiling and tracking

IS-PAC system analyzes typical passenger’s behaviour in the terminal and on the basis of the received data IS-PAC creates profiles of standard behaviour what allows to analytically estimate deviation from typical model of behaviour and to inform the system operator in due time.

IS-PAC system is intergated with AODB and DCS airport systems what allows not only to analyze passenger’s behaviour but to attach tags with his personal and flight data as well.


Passenger traffic control access

Control access of one passenger is a simple task but handling passenger traffic is quite a different matter.

IS-PAC system memorizes arrangement of metal detector frames, introscopes, body scanners and tracks passenger’s proceeding to sterile area only through all pre-flight formalities, with informing system operator about any attempt to infringe the standard procedures. IS-PAC system also detects passenger’s attempt during boarding to bypass security agent who checks boarding passes.

Tracking proceeding through automatic turnstiles

Introduction of automatic turnstiles in the areas of pre-flight inspection or boarding areas brings to the forefront the task of automated control of possible infringements during passengers’ proceeding through the automatic turnstiles. IS-PAC system allows to automatically track the attempts to go through the turnstile in pairs or bypassing the standard procedure of access by a boarding pass. Tha system timely informs the operator about such incidents.

Baggage tracking

IS-PAC system also makes it possible to track proceeding of baggage from the moment of its check-in at the counter and further through conveyor system. HD cameras are installed at special points to be able to detect all important events from the point of view of security:

  • Baggage falling off the conveyor belt
  • Attempts to remove baggage thus infringing technology of baggage inspection
  • Attempts to steal baggage
  • Attempts to add objects into baggage

Situation center

Data of all cameras of IS-PAC system proceed to server cluster where information is analyzed and sent to system operators’ workplaces.

On the screens of situation center operators analyze events which have alerted IS-PAC system and after that operator makes appropriate decision and informs relevant services about further actions.

Technology of analysis of passenger behaviour data

The basis of technology is multifactor statistical analysis of involuntary movements of people combined with machine learning and including algorithms of computer vision. Method of assessment of psycho-emotional state consists of several stages:

The first stage is a stage of analysis of the scene with usage of algorithms of computer vision that is segmentation of the scene into its components: still background of the scene, moving and still objects of the scene which are not background, well-formalized objects such as person’s silhouette, upper body, head, face area, legs. The most important component at this stage is algorithm of tracking or following the object. Providing a stable objects tracking on the sequence of video frames is technologically complicated task because of dense informational contents of video stream as such. To provide real time analysis of video information with the current computation capacity it is necessary to use special methods of optimization. To separate algorithmically the object which is to be tracked and not to lose it in the next video frames taking into account its possible overlapping with obstacles or neighbouring objects information grading into meaningful and meaningless by mathematical method is applied at the first stages of analysis.

At the second stage evaluation of meaningful parameters of already segmented scene is performed: speed of person’s movement, visible body parts, quantity of still states during the analyzed time intervals, amplitudes, frequency of pose change and descriptor set well-known for separate tasks of computer vision: histograms of oriented gradients, SIFT, FAST.

Numerous parameters for description of person’s posture, position and movement are graded with the help of machine learning methods and multidimensional statistical analysis resulted in weight coefficients of these parameters in building a final model.
For realization of decision making algorithm the ensemble of neural networks is used – the set of connectionist models which make decisions by averaging results of separate models performance. Connectionist models training is performed with taking into account a windowed estimate of complex parameters received at previous stages.